Keyword research without a paid subscription

The standard keyword research guide opens by telling you to sign up for a tool. Then it spends most of its length teaching you to operate that tool: filter by volume, sort by difficulty, export to a spreadsheet. It reads like tool documentation because that is largely what it is, and because the people writing it usually sell the tool or earn a commission on it.

Here is the argument this piece makes. The numbers a paid subscription gives you — monthly search volume, a difficulty score out of one hundred — are the part of keyword research you can most afford to go without when you are small. They are estimates, not measurements. Google does not publish real volume to third parties, so every figure you see in a tool is modelled from clickstream panels and Keyword Planner buckets. The difficulty score is a formula somebody chose, usually a function of the link profiles of the current top ten. Neither tells you the thing you actually need to know, which is whether you can plausibly say something better than what already ranks, to a person who wants what you sell.

That part is free. It just takes attention rather than a subscription.

What keyword research is actually for

Two things: understanding demand, and understanding intent. Everything else is bookkeeping.

Demand means knowing which problems people are actively typing about, and in what words. Not what you call the thing — what they call it. A firm that sells "thermal envelope remediation" is invisible to the person searching "cold bedroom above garage".

Intent means knowing what the searcher wants to happen next. The same topic supports wildly different intents. Someone searching "underfloor heating cost" wants a number and a comparison. Someone searching "underfloor heating installers near me" wants a phone call today. If you point one page at both, you satisfy neither, and Google — which ranks pages against the intent it has already observed for that query — will hand the slot to somebody who picked a side. If you are hazy on why that is, how Google ranking works, explained without jargon covers the mechanics.

A volume number tells you nothing about intent. The SERP tells you everything about intent, and the SERP is free.

Free sources, roughly in order of how much they are worth

Search Console, if you have any traffic at all

If your site has been live for a few months and has any impressions, the Performance report in Google Search Console is better than any paid keyword tool, for one reason: it is not modelled. It is a record of queries where Google actually showed your pages, with impressions, clicks and average position.

The queries to look at are not your top ones. Look at queries with impressions and near-zero clicks, sitting somewhere between positions eight and twenty-five. Those are searches where Google already believes you are roughly relevant but not the best answer. That is the shortest distance between where you are and a page that earns clicks. Filter to the last three months, sort by impressions, and read the query list line by line — you will find phrasings you would never have guessed, because they came from your customers rather than from your assumptions.

Search Console also has a limitation worth knowing: it does not show every query. Rare and personally identifying queries are withheld, so the long tail you see is a sample of the long tail that exists.

Autocomplete, People Also Ask and related searches

Type your core term into Google and read the suggestions without pressing enter. Then type it with a trailing space and a letter — "boiler service a", "boiler service b" — and read again. Autocomplete is built from real query patterns, which makes it a demand signal that costs nothing.

People Also Ask boxes give you the question forms. Expanding one usually spawns more, so you can walk a topic outward for as long as you have patience. Related searches at the foot of the page show you the lateral moves searchers make when the first result did not satisfy them.

None of these come with numbers. All of them come with language, which is more useful.

Google Trends, for comparison only

Trends will not tell you how many people search a term. It tells you how one term compares to another over time, and how that varies by region. That is enough to settle real arguments: is it "handyman" or "odd job man" in the UK, is demand for your service seasonal, is the phrase you have built a page around actually declining. Use it to choose between two candidate phrasings, not to size a market.

Forums, communities and question sites

Where do your customers complain? Reddit threads, Facebook groups, trade forums, Amazon reviews of adjacent products, the question tab on a manufacturer's support site. People in these places describe problems in full sentences, before they have compressed them into a search query. This is where you find the objection you did not know existed and the phrase you would never have invented.

Your own inbox and phone calls

This is the source nobody with a subscription bothers with, and it is the best one. Every question a prospect asks before buying is a query someone else is typing into Google right now. Keep a running note for two weeks. Write down the question in the customer's words, not yours. You will end that fortnight with a content list better than anything an export would have produced.

Competitor page titles

Open your three closest competitors and read their page titles and headings. You are not looking for their traffic — you are looking for the topics they thought were worth a dedicated page, and the gaps where nobody has bothered. A site: search in Google over a competitor's domain gives you the whole list in one screen.

Judging difficulty without a difficulty score

Difficulty scores compress a rich page into one number. Reading the SERP gives you the rich page back. Search your target query in an incognito window and work through this:

| What you see | What it means | Worth attacking? | |---|---|---| | Top results are national brands, Wikipedia, government sites | The query is informational and competitive, and links follow it around | Rarely, early on | | Top results are thin affiliate roundups and content farms | Nobody has done the work properly | Often — this is the best signal there is | | Results are forum threads and Reddit posts | Real demand, no good published answer | Yes, if you can write the answer properly | | A local pack sits above the results | Google reads it as local intent | Yes, if you are local — and your Business Profile matters more than your page | | Results answer a different question than you assumed | Your intent guess was wrong | Fix the intent, then reassess | | Every result is a product or category page | Commercial intent — an article will not rank | Only with a commercial page | | The results are old, and the topic has moved | Freshness gap | Yes |

Two more checks. First, does an ad block sit at the top? Advertisers spend money where money is made, so ads are a rough proxy for commercial value — more honest than a modelled volume figure, because someone is paying real money for it. Second, ask whether you can genuinely produce a better page than the weakest result in the top five. If the answer is no, difficulty is high for you regardless of what any score says. If the answer is obviously yes, difficulty is low for you regardless of the same score. Difficulty is relative to the person doing the work; scores pretend it is a property of the query.

A process you can run in an afternoon

1. Empty the head first. Ten minutes, no research. Write every phrase you think a customer might type. This is your bias, on paper, where you can see it.
2. Pull Search Console. Last three months, all queries, sorted by impressions. Copy out everything with impressions and few clicks. Skip this step if the site is new; you will lean harder on the next two.
3. Walk autocomplete. Take your five strongest head terms. Run them through autocomplete, People Also Ask and related searches. Collect the phrasings, not the topics.
4. Mine one community. Pick the single place your customers actually talk. Half an hour reading. Add the problems in their words.
5. Group by intent, not by keyword. Sort everything into four buckets: someone with a problem, someone comparing options, someone ready to buy, someone already a customer. Most phrases collapse into a handful of groups — each group is one page, not one page per phrase.
6. Open the SERP for one representative query per group. Apply the table above. Note the verdict in a word or two.
7. Prioritise. For each group, score two things one to five: how close it sits to money, and how beatable the SERP looks. Sort by the two combined. That is your list.

The output is a prioritised set of pages with a reason attached to each. No subscription generated any part of it.

Worked example: a two-person plumbing business in one town

The head-empty list comes out as "plumber", "emergency plumber", "boiler repair" — all obvious, all dominated by directories and national call-centre operators who buy leads.

Search Console (the site is two years old, with a services page and three blog posts) shows impressions for things nobody in the business wrote down: "boiler pressure dropping overnight", "radiator cold at bottom", "how much to move a radiator", plus a handful of village names next to "plumber" that are not the town the business is registered in.

Autocomplete on "boiler pressure" produces a cluster of related failure descriptions. People Also Ask supplies the question forms — why it happens, whether it is dangerous, whether it needs an engineer. A local Facebook group turns up the same complaint written out at length, plus a recurring worry about being charged a call-out fee for something trivial.

Grouping by intent gives four pages, not fifteen keywords:

• Problem, urgent: losing boiler pressure. SERP is national brands and manufacturer pages — hard to outrank, but it is the top of the funnel for the actual job, and a genuinely local, plain answer that ends with "here is when it needs an engineer and roughly what that costs" has something the national pages structurally will not print.
• Problem, low urgency: cold radiator at the bottom. SERP is forum threads and thin content. Beatable, and worth doing first.
• Comparison: cost to move a radiator. SERP is price-guide sites with wide national ranges. A page with real local prices beats every one of them on usefulness, and price transparency is the single easiest edge a small firm has.
• Ready to buy: plumber plus the neighbouring village names. Local pack territory. The page matters less than the Google Business Profile, service area settings and reviews.

The prioritised order writes itself: cold radiator first because it is winnable, radiator cost second because it is near money, village pages third because the profile work matters more than the page, boiler pressure last because it is a long campaign. That took an afternoon, and it is a better brief than a keyword export because every line has a reason attached.

If the site were brand new rather than two years old, the same process runs — you just lose step two and lean on communities and sales conversations instead. How to promote a business online starting from zero covers what else needs doing at that stage, since keyword research on its own does not move anything.

What you genuinely lose, and when it starts to matter

Be honest about the trade-off. Without a paid subscription you lose two things.

Absolute volume. You cannot say "this term gets roughly four thousand searches a month and that one gets two hundred". You can only say which of two terms is bigger, via Trends, and only for terms with enough volume to register there.

Scale. You cannot process ten thousand keywords in an hour, cluster them automatically, or track hundreds of rankings daily.

Neither matters much when you are choosing your next four pages. Four pages is a decision you make by reading, not by filtering. The precision of a volume estimate is irrelevant when your constraint is that you can publish one good page a fortnight, and when the estimate itself carries a wide error bar.

It starts to matter at three specific points. When you are choosing between markets or building a site plan of hundreds of pages, relative sizing across a large set genuinely needs data you cannot gather by hand. When you are forecasting to someone who controls a budget, "beatable SERP, close to money" is a weaker case than a modelled number, even though the number is softer than it looks. And when you have a team producing volume, the cost of a subscription disappears against the cost of pointing three writers at the wrong topics for a month.

Until then, a tool mostly buys you the feeling of rigour. That feeling is not nothing, but it is not four figures a year either.

Mistakes that cost people the most

• Chasing a phrase whose SERP shows a different intent. The most common failure by a distance. Look before you write.
• One page per keyword. Five phrasings of the same question are one page. Splitting them means five weak pages competing with each other.
• Ignoring the queries already in Search Console in favour of hunting new ones. Existing impressions are the cheapest wins on the site.
• Treating research as a one-off. Your Search Console query list changes as you publish. Re-read it monthly; it will tell you what to write next more reliably than any brainstorm.
• Expecting the research to be the hard part. It is not. Publishing, and then waiting, is. How long SEO takes and what the evidence supports is worth reading before you set expectations with anyone.

Questions people ask

Is Google Keyword Planner enough on its own? It is free with an Ads account, and it gives you bucketed ranges rather than precise figures unless you are spending. Useful as a sanity check on relative size. Less useful than the SERP for everything else.

Do I need the volume number to write a good page? No. You need to know the question, the intent and whether the current answers are weak. A page that genuinely answers a low-volume question converts better than a page that vaguely addresses a high-volume one.

What if the SERP shows I cannot win anything? That is a real result, and it is worth knowing before you spend six months finding out slowly. It usually means the competitive terms are saturated and your gap is narrower and more specific than you hoped — or that search is not the right channel for you at this stage. Either way, the honest cost comparison of paid versus organic is the next thing to read.

Where does a paid visibility marketplace fit into this? Differently, and it is worth being blunt about it. We run a public leaderboard where position is a function of verified spend, so nothing on this page describes how our own ranking works — there is no keyword research involved in it at all. A marketplace listing is a way to be visible while the organic work matures; it is not a substitute for knowing what your customers type or for having a page worth landing on. If the research above tells you there are winnable queries close to money, do that work. The things pay-to-rank cannot do is the honest version of where our model stops.

The research is free. What it costs is an afternoon of paying attention to your own customers, which is the part no subscription has ever been able to sell.